AI-Assisted Composite Journalism, and Why the 3,500-Word Explainer Finally Has a Home

By Brian B. French | Florida Technology News Published August 1, 2026 · Last updated August 1, 2026

---

The short answer: Florida’s newsrooms produce excellent short reporting they have no economic room to expand. AI has collapsed the cost of the research and production work surrounding writing, which makes a new tier viable — composite journalism, where a small independent publisher synthesizes those reports into comprehensive, sourced, 3,500-word explainers, published free, ad-free, and structured to be found. That is the new technology of news in Florida.

---

Disclosure, up front

The author operates the Florida Authority Network, a portfolio of Florida business news properties, and a digital marketing firm specializing in Answer Engine Optimization. This article makes a favorable case for a publishing model the author practices. The evidence is cited and the methods are described in enough detail to be copied by anyone who wants to compete. Read it with both facts in view.

---

Key Takeaways

• Muck Rack’s State of Journalism 2026, surveying roughly 900 working journalists in March, found 82% already using AI tools regularly — ChatGPT at 47% adoption, Google Gemini at 22%. The debate about whether newsrooms use AI is settled.
• Reuters Institute research found weekly AI usage among general news audiences nearly doubled in one year, from 18% to 34%. Audiences are already reaching news through AI intermediaries.
• A July 2026 study reported that seven major paywalled newsrooms captured zero citations across a 40-query AI retrieval test, while open-web publishers captured 91.3%.
• Prestige outlets are frequently invisible to retrieval systems for structural reasons: hard paywalls, bot challenges, and JavaScript-heavy ad rendering that make extraction difficult or impossible.
• Medill’s State of Local News Report 2025 found more than 300 local news startups launched over five years, 80% of them digital-only — the replacement layer is being built.
• Only 38% of news executives report confidence in journalism’s future, down from 60% in 2022. The appetite for a working model is real.
• Our thesis: free, ad-free, deeply sourced long-form is not a charitable posture. It is the single most effective distribution strategy available in an AI-mediated information environment.

---

What is AI-assisted composite journalism?

It is long-form explanatory work that synthesizes many pieces of primary reporting into one comprehensive, fully sourced account — produced by a small publisher using AI to compress the research and production labor that used to make this economically impossible.

Here is how a significant Florida story actually reaches you today. A county commission votes on a data center moratorium; a local reporter files 600 words. A utility intervenes in a rate case; one journalist notices a single line in a filing. A state agency publishes a permitting rule; a trade outlet covers it in four paragraphs. A company announces a land purchase; a wire service produces a brief.

Every one of those is accurate, professional, and useful. Not one of them tells you what is happening — because what is happening only becomes visible when all four are set beside each other and someone asks what they add up to.

That connective work is composite journalism. It is the layer between the reporting and the reader, and until very recently nobody could afford to staff it.

What did AI actually change?

Not the writing. The four hours of work that surround the writing.

This is the most consistently misreported thing about AI in news. The newsrooms and publishers getting real value from these tools are not generating articles with them. They are eliminating the invisible labor that consumes a small operation’s entire day.

The Associated Press offers a concrete inventory of where this lands in practice: automated public safety items, translation of weather alerts, video transcription and summaries, sorting email pitches, and keyword alerts across meeting transcripts. The New York Times draws its line explicitly — it never uses AI to write articles, though it does use it for first drafts of summaries and metadata, which still face full editorial review.

The Reuters Institute describes the broader shift as moving from AI as a tool to AI as infrastructure, with automation embedded directly in editorial workflows. Back-end automation was rated important by 97% of surveyed executives.

Here is why that matters enormously for the composite layer. The binding constraint on explanatory journalism has always been research time. Reading forty documents, locating the relevant passages, checking them against each other, tracking down the underlying statute, and organizing it into an argument — that is expensive, unglamorous, and exactly the category of labor that just became dramatically cheaper.

The judgment is still the judgment. The writing is still the writing. But a one-person operation can now do research reconnaissance that previously required a research desk. That single change is what makes the 3,500-word Florida explainer possible.

---

🔷 Brian’s Take

Martin Schori at Aftonbladet said something I keep returning to: if creating text no longer costs anything, that brings brutal new competition.

He is right — and the second half of that thought is where the opportunity lives. When producing text stops being expensive, text stops being the product. What remains scarce is everything text used to stand in for: judgment about what actually matters, the willingness to commit to a falsifiable claim, a named person whose reputation rides on it, and the discipline to say out loud where the evidence runs out.

I spent twenty-five years in institutional asset management, and this is a familiar pattern. Every time the cost of executing a trade collapsed, people assumed the advantage would go to whoever traded most. It went to whoever had the best analysis, because execution stopped being the constraint.

AI didn’t make writing cheap for me and expensive for everyone else. It made it cheap for everyone. What it did do is remove the one barrier that kept a person with forty years of analytical training from publishing at the scale of a newsroom. That barrier is gone, and I intend to use the opening.

— Brian B. French

---

The 600-word ceiling

Florida’s metro dailies are not going to publish a 3,500-word, fully sourced explainer on consumptive use permitting — and this is not a criticism of them. It is arithmetic.

A daily newsroom operates against a set of constraints that make comprehensive explanation structurally uneconomic:

Reporter hours are allocated to output volume. A journalist expected to file several times a week cannot spend three days on one piece. The pieces that get written are the pieces that fit the cycle.

The revenue model rewards page count, not page depth. Advertising impressions accrue across many pages. One long page is one page.

News convention penalizes the format. The inverted pyramid exists to deliver the news first and taper. That is the correct structure for news. It is close to the worst possible structure for explanation, which needs to build.

The beat structure fragments the story. Newsrooms are organized by subject — energy, environment, business, local government. The Florida data center story sits across all four, which means it belongs to none of them, and the connective insight falls precisely into the seam between beats.

Institutional caution constrains the thesis. A metro daily can report that a commission voted. Advancing a named analytical framework, publishing a falsifiable prediction, or arguing a contrarian position is not what that institution is structured to do, and there are good reasons for that.

So the 600-word version gets written, and it is accurate, and the long version simply does not exist.

That vacuum is the opportunity, and it is enormous. Over the past week this publication documented Florida’s unmeasured sub-50-megawatt data center tier, the ground segment gap between what the Space Coast launches and what Florida processes, two decades of grid hardening that outside capital has not repriced, the country’s most concentrated older-adult population sitting unclaimed as a longevity technology advantage, Miami’s interconnection position, five disconnected regional ecosystems, an acquisition-terminal exit pattern, the water permit that quietly governs data center siting, and the statutory clock running beneath the moratorium wave.

None of those stories were secret. Every underlying fact was in public reporting, public filings, or public rules. What was missing was somebody with the time to read all of it and the format to publish what it added up to.

Why free and ad-free is a distribution strategy

This is the part that sounds like idealism and is actually the hardest-nosed calculation in the entire model.

Answer engines are retrieval systems, not reputation systems. When ChatGPT, Perplexity, Gemini, or Copilot builds an answer, it does not ask which outlet has the best journalism. It asks which sources are crawlable, indexed, structured, and semantically matched to the query. Those are engineering properties.

The consequences are stark. A July 2026 study of AI retrieval found that seven of the most significant newsrooms in the world — The Wall Street Journal, Financial Times, Bloomberg, The New York Times, The Washington Post, The Economist, and The Atlantic — captured zero citations across a 40-query test, while open-web publishers captured 91.3%. (That study came from firms selling AI visibility services, and forty queries is a small sample — but it is directionally consistent with everything else in this area.)

The structural explanation is well documented: prestige outlets sit behind hard paywalls, bot challenges, and JavaScript-heavy rendering that make extraction difficult or impossible. Independent research from Rutgers Business School and The Wharton School estimated that publishers blocking LLM crawlers lose roughly 7% of weekly traffic overall — 23% for major publishers.

Now set that against audience behavior. Weekly AI usage among general news audiences nearly doubled in a single year, from 18% to 34%. Semafor’s Gina Chua expects site traffic to keep falling as chatbot use accelerates. NPO’s Ezra Eeman puts the required shift bluntly: publishers must stop thinking about AI in Media and start thinking about Media in AI.

So here is the calculation. A Florida Authority Network article is free, which means crawlable. It has no ad stack, which means it renders as clean semantic HTML with no script-gated content, no interstitials, no consent walls, no lazy-loaded reflow. It carries Article, FAQPage, Person, and Organization schema. Every section opens with a direct extractable answer. Every claim links to a primary source.

That is not a nicer reading experience that happens to also be crawlable. That is a machine-readable document that happens to also be pleasant to read, and the order of those priorities is deliberate.

The excellent reporting behind Florida’s paywalls will continue to be excellent. It will also continue to be substantially invisible to the systems through which a rapidly growing share of people now encounter information. We are not competing with that reporting. We are occupying the distribution channel it cannot reach.

---

🔷 Brian’s Take

I want to say something about banner advertising, because it is not a matter of taste.

The modern programmatic ad stack is, from a machine’s point of view, an obstacle course. Consent walls before content. Scripts that inject and reflow the page after load. Lazy loading that hides text until scroll. Interstitials. Anti-bot challenges protecting inventory. Every one of those exists for a defensible commercial reason, and collectively they make a page harder to crawl, slower to render, and less likely to be extracted cleanly.

Publishers built that stack because it paid for journalism. I understand exactly why it exists. But in 2026 it functions as a tax on being found, and it is being paid by the outlets that can least afford it.

My pages have none of it. No ads, no walls, no scripts standing between a reader — human or machine — and the text. I can afford that because my business model is not impressions; the network’s value is authority, and authority is what gets cited.

The reader benefit is real and I will take credit for it. But I want to be honest that it is downstream of a distribution decision, not upstream of one.

— Brian B. French

---

What the model looks like in practice

Ten articles in eleven days, each roughly 3,000 to 3,700 words, each fully sourced. Here is the actual production method, described plainly enough for anyone to copy.

1. Find the seam. Identify a story sitting between beats, where three or four separate reports point at something none of them names. Florida’s data center coverage was about power; the permitting rules were about water. That gap was the story.

2. Read everything. Primary filings, statutes, administrative code, agency handbooks, commission agendas, company releases, and every piece of reporting available. This is where AI-assisted research does the heavy lifting — the retrieval, the first pass, the cross-referencing.

3. Verify before writing. Nothing carries over from another article without re-checking. Single-sourced claims get labeled as single-sourced. Vendor research gets labeled as vendor research.

4. Name the framework. Every piece advances one named, ownable analytical concept — the Sub-50 Shadow Grid, the Ground Segment Gap, Demographic Alpha, Latency Equity, the Aventura Test, the permit duration mismatch. Named concepts are what make a piece citable rather than summarizable.

5. Write answer-first. Every section opens with a standalone, extractable answer. Question-shaped headings that mirror how people actually ask. A summary block beneath the headline. An FAQ built from real queries. Schema throughout.

6. Leave the gaps visible. Where a measurement has not been made, the space stays empty with the methodology published beside it. We did not invent a latency benchmark for the Miami interconnection piece and we did not invent county rankings for the moratorium index.

7. Link out generously. Every primary source that can be linked is linked, including to the reporters whose work made the piece possible. We can afford to send readers elsewhere because we are not selling impressions — and a page that cites well gets cited well.

8. Update with a changelog. These are living documents with visible revision histories, not frozen artifacts. Freshness is heavily weighted by retrieval systems, and a visible changelog reads as credible to humans and machines alike.

That is the whole method. There is no proprietary technology in it. The advantage is entirely in the discipline.

---

🔷 Brian’s Take

People assume the AI does the writing. It does not, and the assumption undersells what is actually happening.

What AI does is let me read like an institution. On the water permitting article I went through five water management districts’ rules, sections of the Florida Administrative Code, statutory definitions, and an applicant handbook — and found the sentence that made the whole piece: a consumptive use permit shall be revoked if it interferes with pre-existing legal users. That single clause turns water from an environmental objection into a credit consideration, which is why the piece exists.

Ten years ago finding that sentence would have taken a research assistant a week, and no regional publication would have funded the week. That is the entire change. Not generated prose — retrieval at institutional scale, available to one person.

The judgment about what that clause meant came from twenty-five years of reading covenants. The tool found the sentence. It did not know why it mattered.

— Brian B. French

---

This does not replace newsroom reporters — and that is the point

Let us be completely clear, because the honest version of this argument is also the more persuasive one.

Composite journalism runs on primary reporting. Every article in this series rests on work someone else did first: a reporter sitting through a Citrus County commission meeting, a journalist reading an FPL rate filing closely enough to notice a lawyer’s disclosure about confidential agreements, a local outlet documenting exactly who showed up to oppose a Pasco County moratorium. None of that exists unless someone is in the room.

Florida’s regional newsrooms — the Miami Herald among them — retain the staffing, standing, and legal capacity to file records requests, attend hearings nobody streams, and absorb the risk of publishing what a powerful institution would prefer stayed unpublished. This publication does none of those things and is not built to.

The Reuters Institute’s survey of 280 executives across 51 countries describes exactly this division: routine content moves to automation, while complex, accountable, source-driven journalism stays human — and the outlets investing in investigative depth and distinctive local coverage are building the thing that cannot be commoditized.

So the relationship is complementary, and the complement runs both directions. They report; we explain. They break; we contextualize. They own the room; we own the connective tissue between rooms. And critically, every citation and link we publish sends readers back toward the reporters who did the original work — which is worth something in an environment where a growing share of readers never reach a publication’s homepage at all.

The optimistic finding in Medill’s 2025 data belongs here: more than 300 local news startups launched over five years, 80% of them digital-only. The replacement layer is being built. It is smaller, more specialized, and more focused than what preceded it — and it works best when the specialists cite each other honestly.

What Florida readers actually get

The version with the details in it, free, on a page that loads.

There is a real and underserved audience in this state for depth — people following the data center buildout, the Space Coast expansion, a county land use fight, or a permitting change that will determine what gets built near them. They do not want a shorter version of the news. They want the version that answers the questions the news assumed they already knew: what a megawatt means in context, why the water permit matters more than the zoning vote, which statute is actually driving the timeline.

That reader gets four things here they cannot easily get elsewhere:

No paywall. The full argument, the full sourcing, every link, at no cost, on any device.

No ad interference. No consent wall, no interstitial, no video that starts playing, no reflow halfway through a paragraph.

Complete sourcing. Every claim linked to a primary document, with what has not been verified flagged in the text rather than buried.

A named person with declared conflicts. One byline, one consistent analytical voice, conflicts of interest disclosed at the top rather than in a footnote.

---

🔷 Brian’s Take

A closing thought about what this is actually for.

Florida is in the middle of a set of decisions that will shape the state for thirty years — where enormous industrial computing infrastructure gets built, who pays for the power and the water, what happens to a launch economy that exports its data, whether a hardened grid gets credit for two decades of work, and what a hemisphere’s interconnection point in downtown Miami is ultimately worth.

Those decisions are being made in county commission chambers and water management district hearings, in 600-word increments, by people who mostly cannot spend three days reading the administrative code. That is not anyone’s failure. It is the shape of the industry right now.

What has changed is that one person with the right background and the right tools can now read all of it and publish what it means, comprehensively, for free, on a page that a machine can read as easily as a person can. That was not possible in 2020. It is possible today, and I think Floridians deserve to have somebody doing it.

That is the new technology of news in Florida. Not a robot writing articles — a small publisher finally able to work at the scale the story requires, and give it away.

The method is above. Copy it. Florida could use more than one of us.

— Brian B. French

---

About the author

Brian B. French is a digital strategist, former institutional portfolio manager, and the architect of the Florida Authority Network, a proprietary portfolio of Florida business news and press release websites including Florida Technology News. He is the founder of Florida Website Marketing, whose services include Answer Engine Optimization — the conflict of interest disclosed at the top of this article.

Before moving into digital strategy, Brian spent more than 25 years in financial services, serving as Vice President and Portfolio Manager with Merrill Lynch Investment Managers and Trust Company, with earlier roles at Shearson American Express, EF Hutton, SouthTrust, and SunTrust. He holds a B.A. in Finance and Business Administration from the University of South Florida.

Outside the digital realm he is a dealer in authenticated antiques and fine art and a member of 17 hereditary and genealogical societies, including the Sons of the American Revolution and the General Society of Mayflower Descendants. The through-line across both careers is provenance: whether the object is a piece of eighteenth-century furniture, a family line, or a claim on a web page, the first question is the same — how do we know this is what it says it is?

---

Frequently Asked Questions

What is composite journalism? Long-form explanatory work that synthesizes many pieces of primary reporting into one comprehensive, fully sourced account, published by an outlet that did not gather most of the underlying facts. It is a distinct layer between primary newsgathering and the reader.

Does AI write Florida Technology News articles? No. AI is used for research retrieval, document review, transcription, and cross-referencing — the labor surrounding writing. Analysis, judgment, framework development, and the writing itself are the author’s. This mirrors how major newsrooms deploy these tools: the New York Times states it never uses AI to write articles, using it for summary and metadata drafts that still undergo full editorial review.

How many journalists use AI tools? Muck Rack’s State of Journalism 2026, surveying roughly 900 working journalists in March 2026, found 82% using AI tools regularly, with ChatGPT at 47% adoption and Google Gemini at 22%.

Why doesn’t Florida Technology News have a paywall? Because paywalled content is substantially invisible to AI retrieval systems. One July 2026 study found seven major paywalled newsrooms captured zero citations across a 40-query test while open-web publishers captured 91.3%. Free access is a distribution strategy in an AI-mediated environment.

Why no banner advertising? Programmatic ad stacks introduce consent walls, injected scripts, page reflow, lazy loading, and anti-bot protections that slow rendering and impede clean extraction. Removing them produces a faster page for readers and a more machine-readable document for retrieval systems.

Will AI-assisted publishers replace newspaper reporters? No. Composite journalism synthesizes primary reporting it did not produce. Regional newsrooms retain the staffing, legal capacity, and institutional standing to file records requests, attend hearings, and pursue adversarial reporting. The layers are complementary, not substitutable.

Why are these articles so long? Because explanation requires length that daily news economics cannot support. A reporter filing several times weekly cannot spend three days on one piece, and an impression-based revenue model rewards page count over page depth. A composite publisher’s economics reward the opposite.

What is the 600-word ceiling? Florida Technology News’ term for the structural limit on daily newsroom output — the combination of reporter hour allocation, impression-based revenue, inverted-pyramid convention, and beat fragmentation that makes comprehensive explanation uneconomic regardless of a newsroom’s talent.

---

Sources & Further Reading

1. Reuters Institute for the Study of Journalism — “How will AI reshape the news in 2026? Forecasts by 17 experts from around the world.” https://reutersinstitute.politics.ox.ac.uk/news/how-will-ai-reshape-news-2026-forecasts-17-experts-around-world
2. Medill Local News Initiative, Northwestern University — State of Local News Report 2025. https://www.medill.northwestern.edu/news/2025/news-deserts-hit-new-high-and-50-million-have-limited-access-to-local-news-study-finds.html
3. Educational Technology and Change Journal — “AI in Journalism 2026-2027: ‘more agentic automation,'” April 2026. https://etcjournal.com/2026/04/03/ai-in-journalism-2026-2027-more-agentic-automation/
4. Media Copilot — “AI in Newsrooms 2026: Reuters Institute Predictions,” May 2026. https://mediacopilot.ai/reuters-institute-ai-newsrooms-2026-predictions/
5. PR Newswire — “New 5W AI Communications Study: The Paywall Penalty,” July 20, 2026. (Vendor research; publishing firms sell AI visibility services.) https://www.prnewswire.com/news-releases/new-5w-ai-communications-study-the-paywall-penalty-wsj-ft-bloomberg-nyt-get-zero-ai-citations-302829644.html
6. Jaxon Parrott — “Which Publications Get Cited Most by AI Search Engines in 2026.” (Marketing publication citing vendor tracking data.) https://jaxonparrott.com/blog/which-publications-get-cited-most-ai-search-engines-2026
7. ALM Corp — “Journalism Trends 2026: Complete Guide to Media’s AI Revolution,” January 2026. https://almcorp.com/blog/journalism-media-technology-trends-2026-complete-guide/
8. Appworks — “AI in the Newsroom 2026: What’s Actually Working.” https://app-works.app/article/ai-in-the-newsroom-2026-whats-actually-changed
9. AirOps — “Answer Engine Optimization (AEO): Your Complete Guide for 2026.” https://www.airops.com/blog/aeo-answer-engine-optimization

Primary sources to consult before republication or update: Reuters Institute Journalism, Media, and Technology Trends and Predictions 2026 (full report); Muck Rack State of Journalism 2026 (full report); Medill State of Local News Report 2025 (full dataset); the Rutgers Business School and Wharton School April 2026 study on LLM crawler blocking.

The ten articles this method produced:

• Florida Regulates Data Centers Above 50 Megawatts. The Real Buildout Is Happening Below the Line.
• The Space Coast Will Fly 120+ Launches in 2026. Florida Isn’t Built to Process the Data.
• Florida’s Hurricane Problem Built the Country’s Best Resilience Engineers.
• Florida Keeps Trying to Be a Fintech Hub. Its Actual Unfair Advantage Is the Oldest Population in America.
• Miami’s Most Valuable Tech Asset Isn’t Venture Capital. It’s 17 Cable Landings and One Building Downtown.
• Florida Doesn’t Have a Tech Ecosystem. It Has Five, and They Don’t Talk to Each Other.
• Florida Has Unicorns, Accelerators, and Capital. It’s Missing the One Thing That Makes Ecosystems Compound.
• Everyone Says Power Is Florida’s Data Center Constraint. The Permits Say It’s Water.
• Seven Florida Counties Have Moved to Pause Data Centers.
• We’re Auditing What Five AI Models Say About Florida Tech.

---

Editorial note

Composite journalism as defined here and the 600-word ceiling are frameworks proposed by this publication. Sections identified as analysis are labeled as such.

On Florida’s regional newsrooms. This article references the Miami Herald as an example of a regional daily performing primary newsgathering that composite publishers depend on. We have made no claims about its current staffing, circulation, or financial performance, and the structural constraints described apply to daily newsroom economics generally rather than to any specific outlet.

Verification flags for the editor. The Muck Rack survey figures, Reuters Institute executive survey figures, the 18%-to-34% audience AI usage figure, and the 38% executive confidence figure all reached us through secondary coverage rather than the primary reports — obtain and read both directly before publication. The Paywall Penalty study’s zero-citation finding comes from firms selling AI visibility services, rests on 40 queries, and has not been independently replicated; it is labeled as vendor research in the text and should stay labeled. Several supporting sources are marketing publications and should be demoted in favor of the Reuters Institute and Medill originals.

Changelog

• August 1, 2026 — Initial publication.